Risk of diabetes in a real-world setting among patients initiating antihypertensive therapy with valsartan or amlodipine

Article metrics


In the Valsartan Antihypertensive Long-Term Use Evaluation (VALUE) trial, the risk of new-onset diabetes was reported to be 23% lower among patients initiating therapy with valsartan versus amlodipine. The objective of our study was to examine whether this finding is generalizable to ‘real-world’ clinical practice. A retrospective cohort design and a large US health insurance database were employed for analyses. Study subjects included all hypertensive patients, aged 35 years, who were free from diabetes and who initiated treatment with valsartan (n=9999) or amlodipine (n=18 698) between January 1999 and March 2005. Unadjusted absolute risks of diabetes were 21.4 (95% confidence interval (CI) 18.9–24.3) and 26.3 (95% CI 24.3–28.3) per 1000 patient-years for valsartan and amlodipine, respectively; the corresponding relative risk (RR) for valsartan was 0.82 (95% CI 0.70–0.94). Multivariate analyses – controlling for age, sex, presence of hypercholesterolemia, cardiovascular disease and kidney disease, and pretreatment medical care expenditures – yielded similar results (RR=0.79, 95% CI 0.68–0.92). Our study thus corroborates the finding from VALUE that diabetes risk is lower for patients who receive valsartan versus amlodipine, and extends this finding to a ‘real-world’ setting.


Approximately 65 million Americans, or about one in three adults, suffer from hypertension.1 In addition to the well-known risks of cardiovascular disease (CVD) conferred by hypertension, persons with this condition also are more likely to develop type II diabetes, which is itself an important independent risk factor for CVD (as well as other diseases).2, 3, 4, 5, 6, 7 Because the coexistence of hypertension and diabetes has a profound impact on morbidity and mortality, much attention has been focused on strategies aimed at reducing the incidence of diabetes among persons with elevated blood pressure.3, 7

Recent evidence suggests that the choice of antihypertensive therapy may be one important component of a comprehensive strategy to reduce the risk of diabetes in this patient population. In a systematic review of relevant studies, Padwal et al.8 concluded that existing data ‘suggest that type II diabetes incidence is unchanged or increased by thiazide diuretics and beta-blockers,’ and that ‘the incidence of type II diabetes appears unchanged or decreased by ACE inhibitors, calcium channel blockers, or angiotensin receptor blockers.’ Moreover, a meta-analysis of recent clinical trials of inhibitors of the renin–angiotensin system, including ACE inhibitors and angiotensin-receptor blockers, suggests that such therapy may significantly reduce the risk of new-onset diabetes in hypertensive patients.9 It should be noted, however, that this analysis did not account for differences across trials in comparators (often placebo).

Recently, the Valsartan Antihypertensive Long-term Use Evaluation (VALUE) trial was the first large randomized trial to compare the risks of diabetes between patients treated with an angiotensin-receptor blocker versus a calcium antagonist, an agent believed to have neutral metabolic effects.10 In this double-blind clinical trial, 15 245 hypertensive patients at high risk of cardiac events were randomly assigned to treatment with valsartan or amlodipine in a stepped-care regimen that permitted concomitant therapy with hydrochlorothiazide in intermediate steps, and other antihypertensives in the final step. New-onset diabetes was a prespecified secondary end point for the subgroup without this condition at baseline. Over a mean follow-up of 4.2 years, 11.5% of valsartan patients and 14.5% of those receiving amlodipine developed new-onset diabetes; the corresponding odds ratio was 0.77 (95% CI: 0.69–0.87) in favor of valsartan.11

Persons participating in clinical trials, however, typically represent a carefully selected subset of patients seen in clinical practice with the condition of interest; in the VALUE trial, the study population consisted of hypertensive patients at particularly high risk of cardiac events. Moreover, medication use in clinical trials often differs from that in the real world.12 Accordingly, effects reported in clinical trials may differ substantially from those observed in typical clinical practice. We thus undertook an examination of the risk of diabetes among hypertensive patients initiating therapy with valsartan or amlodipine in a real-world setting to explore the generalizability of findings from the VALUE trial.

Materials and methods

Data source

Data were obtained from a large US health insurance database (PharMetrics Patient-Centric Database PharMetrics, Inc., Watertown, MA, USA), and spanned the period from January 1, 1998 to March 31, 2005. The database is comprised of paid facility, professional-service and retail (i.e., outpatient) pharmacy claims from approximately 50 US health plans. These plans provide commercial health maintenance organization, preferred provider organization, Medicare Risk and indemnity products to 15 million persons annually. Approximately 4% of all plan members in the database are aged 65 years or older. Plan members reside throughout the US (Midwest: 35%, Northeast: 18%, South: 33%, West: 14%).

Data available for each facility and professional-service (i.e., medical) claim include dates and place of service, diagnoses (in International Classification of Diseases, Ninth Edition, Clinical Modification (ICD-9-CM) format), procedures performed/services rendered (in Health Care Financing Administration Common Procedure Coding System format), quantity of services provided and provider specialty. Data available for each retail pharmacy claim include the drug dispensed (in National Drug Code (NDC) format), the dispensing date, quantity dispensed and number of days of therapy supplied. All medical and pharmacy claims include amounts charged by health-care providers for goods and/or services provided as well as amounts paid (i.e., reimbursed) by the health plan. Demographic and eligibility data for members are also available and include age, sex, geographic region, coverage type and the dates of eligibility for plan benefits.

All patient identifiers in the database have been fully encrypted, and the database is fully compliant with the Health Insurance Portability and Accountability Act of 1996.13 Since ‘… subjects cannot be identified, directly or through identifiers linked to the subjects …’ (45 CFR 46 §46.101), institutional review board approval for this study was not sought.14

Study population

The study population consisted of hypertensive patients, aged 35 years and older, who were free from diabetes and initiated monotherapy with valsartan or amlodipine between January 1, 1999 and March 31, 2005. Patients were considered to have initiated valsartan or amlodipine therapy if their first prescription for one of these agents was preceded by a 12-month period of continuous medical and drug benefits (‘pretreatment period’) without evidence of receipt of amlodipine, valsartan, or any other antihypertensive medication, including α-blockers, angiotensin-converting enzyme inhibitors, angiotensin-receptor blockers, β-blockers, calcium channel blockers and diuretics.

The presence of hypertension was identified on the basis of one or more medical claims with a corresponding diagnosis code (ICD-9-CM 401-404.9) during the pretreatment period (or on the date of therapy initiation). The presence of diabetes was ascertained during the pretreatment period (or on the date of therapy initiation) based on one or more medical claims with a corresponding diagnosis code (ICD-9-CM 250.x) or one or more prescriptions for an antidiabetic medication. The use of antihypertensive and antidiabetic medications was determined based on pharmacy claims with corresponding NDC codes.

Study outcome

Study subjects were followed for new-onset diabetes from the day after therapy initiation through the date of disenrolment from the health plan (due to death or other reasons) or the end of the study database (March 31, 2005), whichever occurred first. An ‘intent-to-treat’ approach was thus employed in which all subjects were tracked over time for onset of diabetes regardless of contemporaneous events, such as non-compliance with study drugs and changes in antihypertensive therapy (e.g., change in drug dose, addition of hydrochlorothiazide). Onset of diabetes was ascertained on the basis of two or more outpatient diagnoses (7 days apart), one or more inpatient diagnoses, or one or more prescriptions for an antidiabetic medication. For persons meeting one or more of these criteria, the date of the first such encounter was considered to be the date of disease onset.

Statistical analyses

Baseline demographic and clinical characteristics of patients in each treatment group were examined, including age, sex, comorbid conditions (i.e., the presence of hypercholesterolemia, CVD and kidney disease), numbers of outpatient encounters and hospitalizations, and total medical care expenditures; duration of therapy with study drug and duration of follow-up were also examined. Age was assessed as of the date of initiation of valsartan or amlodipine therapy, whereas comorbid conditions and measures of utilization and expenditures were assessed during the 12-month pretreatment period. Comorbid conditions were selected for inclusion based on their hypothesized association with the risk of diabetes and the reliability with which they could be identified in the study database. The presence of the comorbid conditions was ascertained on the basis of two or more outpatient diagnoses (7 days apart), one or more inpatient diagnoses, or one or more prescriptions for pharmacotherapy, as appropriate. Statistical comparisons of patients in the two treatment groups were undertaken using an independent-samples t-test for continuous measures and a χ2-test for discrete ones.

The relative risk (RR) of diabetes per 1000 patient-years for those initiating therapy with valsartan versus amlodipine was estimated along with corresponding 95% CI (based on non-parametric bootstrapping). Kaplan–Meier methods were employed to depict unadjusted (crude) time to the onset of diabetes; a two-sided log-rank statistic was used to compare risks between the treatment groups. A Cox proportional hazards model was used to estimate the RR of diabetes for valsartan versus amlodipine controlling for potential confounders; all patient characteristics noted above were entered into and retained in the model, with the exception of measures of health-care utilization (as they are highly correlated with heath-care expenditures, which was included in the model). Patients who did not develop diabetes prior to the end of follow-up were censored as of the corresponding date (i.e., date of disenrollment or end of study database). The appropriateness of the proportional hazards assumption for the treatment group variable was assessed by including a time-dependent covariate in the model.


Patient characteristics

A total of 530 072 persons in the study database were aged 35 years and had one or more prescriptions for valsartan (n=135 324) or amlodipine (n=394 748) at anytime during the period of observation (Figure 1). Of these patients, 28 697 initiated monotherapy with valsartan (n=9999; 7.4% of total) or amlodipine (n=18 698; 4.7% of total) during the period of interest were free from diabetes at baseline, and met all other sample selection criteria, and thus were eligible for analyses.

Figure 1

Selection of study population.

Valsartan patients were 1 year younger, on average, than those receiving amlodipine (51 versus 52 years, P<0.01) and men constituted a higher proportion of the valsartan group (55 versus 51%, P<0.01) (Table 1). Although the prevalence of selected comorbidities varied between treatment groups, similar percentages of valsartan and amlodipine patients were free from all of the selected comorbidities (80.1 and 80.8%, respectively, P=0.16). Most patients in the study population (valsartan: 79%; amlodipine: 68%) initiated treatment with study drug during 2002 or subsequently.

Table 1 Characteristics of hypertensive patients without diabetes initiating therapy with valsartan or amlodipinea

Risk of diabetes

Diabetes developed in 239 valsartan patients over a mean follow-up of 407 days and 619 amlodipine patients over a mean follow-up of 460 days. Corresponding unadjusted absolute risks of diabetes were 21.4 (95% CI 18.9–24.3) and 26.3 (95% CI 24.3–28.3) per 1000 patient-years, and the corresponding RR was 0.82 (95% CI 0.70–0.94) in favor of valsartan (Figure 2). In multivariate analyses, the RR of new-onset diabetes for patients receiving valsartan compared with those receiving amlodipine was 0.79 (95% CI 0.68–0.92) (Table 2). Tests of proportionality suggested that the RR of diabetes with valsartan versus amlodipine was constant over time.

Figure 2

Unadjusted absolute risk of diabetes among hypertensive patients initiating therapy with valsartan or amlodipine.

Table 2 Cox proportional hazards model for risk of diabetes among hypertensive patients initiating therapy with valsartan or amlodipine


The VALUE trial demonstrated that patients with hypertension who received valsartan therapy were at lower risk of new-onset diabetes than those who were treated with amlodipine. The generalizability of this finding to typical clinical practice was unclear, however. The current study extends this result to a real-world setting. Whereas the biological basis for this apparent benefit of valsartan therapy is uncertain at this time, the most likely reason is a favorable effect of valsartan on glucose metabolism due to interference with the renin–angiotensin system (calcium antagonists are generally regarded to be neutral in their effects on glucose metabolism).15, 16, 17

The risk reductions with valsartan therapy observed in the present study are quite similar to those reported in the VALUE trial, despite some important differences in study designs and study populations. First, we ascertained the onset of diabetes based on diagnoses and prescription medications, whereas in the VALUE trial, ascertainment was based on adverse event reporting, reports of new antidiabetic medication and/or increased fasting blood glucose at study end; the current analysis therefore might have missed cases that would have been detected in a trial setting. This difference might explain, at least partially, why the incidence of diabetes in the current study (2–3% annually) was somewhat lower than that in the VALUE trial (3-4% annually) and other studies (3% annually).4, 10, 11 Second, subjects in this study were, on average, 15 years younger than those in VALUE. This disparity is undoubtedly due to a difference in the minimum age for study entry (35 years in the current study versus 50 years in VALUE) as well as the fact that our study employed data from private health plans insuring a mostly non-elderly population. Third, patients in the VALUE trial had to be at relatively high risk of cardiovascular events, based on an algorithm that included age, sex, cardiovascular risk factors and pre-existing CVD. There was no such requirement in the present study, as the objective was to focus on a broad population of hypertensive patients in real-world clinical practice. Furthermore, more than 92% of the patients in VALUE had been previously treated for hypertension, whereas subjects in the current study were initiating treatment for hypertension. Fourth, patients in the current study received valsartan or amlodipine as monotherapy, whereas in the VALUE trial, the majority of patients received additional antihypertensive drugs, most commonly diuretics. The potentially confounding effects of concomitant use of other antihypertensive medications at baseline were thus minimized in the current study. (It should be noted that use of other antihypertensive medications following initiation of study drugs was permitted, and a sizable proportion of patients did receive such therapy at some time during follow-up, most notably diuretics (21% in both groups), β-blockers (valsartan patients, 7%; amlodipine patients, 11%) and angiotensin-converting enzyme inhibitors (valsartan patients, 5%; amlodipine patients, 17%); in addition, 6% of valsartan patients received calcium-channel blockers and 8% of amlodipine patients received angiotensin-receptor blockers. However, the confounding effect of the use of other antihypertensive medications during follow-up was probably minimal, as the RR of diabetes with valsartan versus amlodipine was largely unaffected (hazard ratio=0.81, 95% CI: 0.66–1.01) when patients were censored on the date of change in, or discontinuation of, their therapeutic regimen. Because neither patients adding valsartan or amlodipine to established regimens nor those switching to these drugs from other antihypertensive agents were included in this study population, the findings cannot be generalized to these groups. Finally, compliance with study drugs was poor – as it typically is in clinical practice for treatments of asymptomatic chronic conditions – resulting in patients having supply of study drug for only about 50% of their follow-up days, on average. To address the possibility that study results may not necessarily be reflective of the effects of study drugs but rather other factors, we undertook additional analyses focusing on subgroups of patients with supply of study drug covering a minimum of 90, 180, 360 and 720 days, alternatively. In all such analyses, the estimated RR of diabetes with valsartan versus amlodipine was generally consistent with our basecase estimate.

Several additional comments on study design are necessary. First, as noted above, the analysis almost certainly missed some persons who may have developed diabetes following therapy initiation but whose disease was not clinically recognized. Second, whereas we were able to control for differences between treatment groups in observed patient characteristics – such as age, sex and comorbid conditions – in multivariate analyses, it is possible that unobserved characteristics (e.g., blood glucose, family history, body mass index) or significant risk factors that have not as yet been identified in the medical literature were not evenly distributed between the treatment groups, and therefore could have confounded study results. For example, if patients who received amlodipine were more likely to be overweight or obese, to have higher fasting blood glucose levels, or to have clinically recognized diabetes than those receiving valsartan, then the estimates of the RR of diabetes with valsartan would have been downwardly biased. It is highly unlikely, however, that patients who initiate treatment with valsartan would be inherently less likely to develop diabetes versus those beginning treatment with amlodipine, or less likely to have clinically recognized diabetes – especially since these two drugs were not believed to differ in diabetogenic effect before the publication of findings from the VALUE trial in 2004, and since the two treatment groups were quite similar in terms of their observed characteristics. Although many small differences in patient characteristics between the two treatment groups were statistically significant as a result of our large sample size, it is highly doubtful that any of these differences are clinically relevant. Moreover, in analyses focusing on patients receiving valsartan or amlodipine who were free from all comorbid conditions of interest at therapy initiation, the RR of diabetes with valsartan was found to be similar (hazard ratio=0.76, 95% CI 0.64–0.90) to our ‘basecase’ estimate. Thus, the effects of confounding due to selection bias or case ascertainment bias are probably limited. Third, we focused attention on patients aged 35 years and older as the prevalence of hypertension – and thus use of valsartan and amlodipine – is relatively low among younger adults, as is diabetes risk. Finally, by employing an ‘intent-to-treat perspective’, actions during follow-up that may affect diabetes risk, such as non-compliance with study drug and change in antihypertensive therapy, were implicitly disregarded. To the extent that they differed systematically between treatment groups, our comparison of diabetes risk for valsartan and amlodipine may be biased. However, as noted above, results of additional analyses undertaken to address this issue (i.e., analyses that censored patients on the date of change in, or discontinuation of, study therapy, and analyses that focused on patients who received study drug for an extended period of time) were generally consistent with our basecase estimate.

In conclusion, this study confirms findings from the VALUE trial that patients with hypertension who initiate treatment with valsartan are less likely to develop diabetes than those who receive amlodipine. The magnitude of risk reduction is similar in this retrospective cohort study and the prior randomized controlled trial, despite significant differences in study designs and patient populations. Although the precise biological basis for this difference in diabetes risk has not yet been well characterized, the analysis appears to confirm its existence in typical clinical practice.


  1. 1

    Fields LE, Burt VL, Cutler JA, Hughes J, Roccella EJ, Sorlie P . The burden of adult hypertension in the United States: 1999–2000. Hypertension 2004; 44: 398–404.

  2. 2

    Chobanian AV, Bakris GL, Black HR, Cushman WC, Green LA, Izzo JL et al. Seventh report of the joint national committee on prevention, detection, evaluation, and treatment of high blood pressure. Hypertension 2003; 42: 1206–1252.

  3. 3

    Kannel WB . Risk stratification in hypertension: new insights from the Framingham Study. Am J Hypertension 2000; 13: 3S–10S.

  4. 4

    Gress TW, Nieto FJ, Shahar E, Wofford MR, Brancati FL . Hypertension and antihypertensive therapy as risk factors for type II diabetes mellitus. N Engl J Med 2000; 342: 905–912.

  5. 5

    Kannel WB . Blood pressure as a cardiovascular risk factor. JAMA 1996; 275: 1571–1576.

  6. 6

    Bonora E, Kiechl S, Willeit J, Oberhollenzer F, Egger G, Meigs JB et al. Population-based incidence rates and risk factors for type II diabetes in white individuals. Diabetes 2004; 53: 1782–1789.

  7. 7

    Verdecchia P, Reboldi G, Angeli F, Borgioni C, Gattobigio R, Filippucci L et al. Adverse prognostic significance of new diabetes in treated hypertensive subjects. Hypertension 2004; 43: 963–969.

  8. 8

    Padwal R, Majumdar SR, Johnson JA, Varney J, McAlister FA . A systematic review of drug therapy to delay or prevent type II diabetes. Diabetes Care 2005; 28: 736–744.

  9. 9

    Jandeleit-Dahm KA, Tikellis C, Reid CM, Johnston CI, Cooper ME . Why blockade of the renin–angiotensin system reduces the incidence of new-onset diabetes. J Hypertens 2005; 23: 463–473.

  10. 10

    Julius S, Kjeldsen SE, Weber M, Brunner HR, Ekman S, Hansson L et al. Outcomes in hypertensive patients at high cardiovascular risk treated with regimens based on valsartan or amlodipine: the VALUE randomised trial. Lancet 2004; 363: 2022–2031.

  11. 11

    Kjeldsen SE, Julius S, Mancia G, McInnes GT, Hua T, Weber MA et al. Effects of valsartan compared to amlodipine on preventing type II diabetes in high risk hypertensive patients: the VALUE trial. J Hypertens 2006; 24: 1405–1412.

  12. 12

    Cardinal H, Tahami Monfared AA, Dorais M, LeLorier J . A comparison between persistence to therapy in ALLHAT and in everyday clinical practice: a generalizability issue. Can J Cardiol 2004; 20: 417–421.

  13. 13

    Health Insurance Portability and Accountability Act of 1996, 42 USC §1320-d2. Available at: http://www.cms.hhs.gov/hipaageninfo/downloads/hipaalaw.pdf(accessed September 5, 2006).

  14. 14

    Public Welfare – Protection of Human Subjects; 45CFR 46 §46.101. Available at: http://ohsr.od.nih.gov/guidelines/45cfr4(accessed September 5, 2006).

  15. 15

    Julius S, Gudbrandsson T, Jamerson KA, Tariq SS, Andersson O . The hemodynamic link between insulin resistance and hypertension. J Hypertens 1991; 9: 983–986.

  16. 16

    Carlsson P-O, Berne C, Jansson L . Angiotensin II and the endocrine pancreas: Effects on islet blood flow and insulin secretion in rats. Diabetologia 1998; 41: 127–133.

  17. 17

    Sharma AM, Janke J, Gorzelniak K, Engeli S, Luft FC . Angiotensin blockade prevents type II diabetes by formation of fat cells. Hypertension 2002; 40: 609–611.

Download references


Funding for this research was provided by Novartis Pharma AG.

Author information

Correspondence to D Weycker.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Weycker, D., Edelsberg, J., Vincze, G. et al. Risk of diabetes in a real-world setting among patients initiating antihypertensive therapy with valsartan or amlodipine. J Hum Hypertens 21, 374–380 (2007) doi:10.1038/sj.jhh.1002159

Download citation


  • diabetes mellitus
  • angiotensin II type 1 receptor blockers
  • calcium channel blockers
  • valsartan
  • amlodipine

Further reading